The ABC of Computational Text Analysis

#1 Introduction +
Where is the digital revolution?

Alex Flückiger

Faculty of Humanities and Social Sciences
University of Lucerne

March 3, 2023

Outline

  1. digital revolution or hype?
  2. about us
  3. goals of this course

AI: A non-standard Introduction

The world has changed, hasn’t it?

An Era of Big Data + AI

Group Discussion

What makes a computer looking intelligent?

AI is a moving target with respect to …

  • human capabilities
  • technological abilities

Transfer of Human Intelligence

from static machines to more flexible devices

  • mimicking intelligent behavior
    • reading + seeing + hearing
    • speaking + writing + drawing
  • a sense of contextual perception
  • many degrees of freedom

Seeing like a Human?

An image segmentation with Facebook’s Detectron2 (Wu et al. 2019)

Speaking like a Human?

Speech-to-Text (STT)

Recognizing speech regardless of language, accent, speed, noise etc.

Check out samples of Whisper (Radford et al. 2022)

Text-to-Speech Synthesis (TTS)

Personalizing voice given an audio sample of 3s

Check out samples of VALL-E (Wang et al. 2023)

Outsmarting Humans?

ChatGPT is amazing but …

… it is also a stochastic parrot. 🦜

(Bender et al. 2021)

Can you disenchant ChatGPT?

Experiment with ChatGPT

  • What works (surprisingly) well?
  • Where does it fail?

Beyond Perception and Unimodality

Generated Images by a Neural Network

https://thisxdoesnotexist.com/

Give me more!

Trend towards Multimodality

Breakthrough by combining language processing and image generation with Muse (Chang et al. 2023)

Deepfakes? It is real!

Editing pictures with Muse using natural language (Chang et al. 2023)

But videos are still real?

Checkout this demo trailer for authentic dubbing.

🎥

Artificial Intelligence

Subfields

  • Natural Language Processing (NLP)
  • Computer Vision (CV)
  • Robotics

How does Computer Intelligence work?

  • interchangeably (?) used concepts
    • Artificial Intelligence (AI), Machine Learning (ML), Deep Learning (DL)
  • learn patterns from lots of data
    • more recycling than genuine intelligence
    • theory agnostically
  • supervised training is the most popular
    • learn relation between input and output

AI Hype in a Nutshell

AI = from humankind import solution

This is how current AI looks like

Why this matters for
Social Science

Computational Social Science

data-driven research

Group Discussion

What kind of data is there?

What data is relevant for social science?

  • data as traces of social behaviour
    • tabular, text, image
  • datafication
    • sensors of smartphone, digital communication
  • much of human knowledge compiled as text

About the Mystery of Coding

coding is like…

  • cooking with recipes
  • superpowers

Women have coding powers too!

Where the actual Revolution is

Coding is a superpower

  • flexible
  • reusable
  • reproducible
  • inspectable
  • collaborative

… to tackle complex problems on scale

About us

Personal Example

directed country mentions in UN speeches

Goals of this Course

What you learn

  • computationally analyze, interpret, and visualize texts
    • command line + Python
  • digital literacy + scholarship
  • problem-solving capacity

Learnings from previous Courses

  • too much content, too little practice
  • programming can be overwhelming
  • learning by doing, doing by googling

Levels of Proficiency

  1. awareness of today’s computational potential
  2. analyzing existing datasets
  3. creating + analyzing new datasets
  4. applying advanced machine learning

What I teach

  • computational practises
  • critical perspective on technology
  • lecture-style introductions
  • hands-on coding sessions
  • discussions + experiments in groups

Topics

Techniques

  • text processing
  • extracting and aggregating information
  • creating simple visualizations
  • optical character recognition (OCR)
  • scraping files

Data

  • using existing datasets
  • creating new datasets


🤓 inputs are more than welcome!

Provisional Schedule

TODO

TL;DR 🚀

You will be tech-savvy…

…yet no programmer applying fancy machine learning

Requirements

  • no technical skills required
    • self-contained course
  • laptop (macOS, Win10, Linux) 💻
    • update system
    • free up at least 15GB storage
    • backup files

Grading ✍️

  • 3 exercises during semester
    • no grades (pass/fail)
  • mini-project with presentation
    • backup claims with numbers
    • work in teams
    • data of your interest
  • optional: writing a seminar paper
    • in cooperation with Prof. Sophie Mützel

Organization

Who are you?

Please fill out this questionnaire

📝

Questions?

Reading

Required

Lazer, David, Alex Pentland, Lada Adamic, Sinan Aral, Albert-László Barabási, Devon Brewer, Nicholas Christakis, Noshir Contractor, James Fowler, Myron Gutmann, Tony Jebara, Gary King, Michael Macy, Deb Roy, and Marshall Van Alstyne. 2009. “Computational Social Science.” Science 323(5915):721–23.

(via OLAT)

Optional

Graham, Shawn, Ian Milligan, and Scott Weingart. 2015. Exploring Big Historical Data: The Historian’s Macroscope. Open Draft Version. Under contract with Imperial College Press.

online

References

Bender, Emily M., Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. 2021. “On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜.” In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, 610–23. Virtual Event Canada: ACM. https://doi.org/10.1145/3442188.3445922.
Chang, Huiwen, Han Zhang, Jarred Barber, A. J. Maschinot, Jose Lezama, Lu Jiang, Ming-Hsuan Yang, et al. 2023. “Muse: Text-To-Image Generation via Masked Generative Transformers.” arXiv. https://doi.org/10.48550/arXiv.2301.00704.
Graham, Shawn, Ian Milligan, and Scott Weingart. 2015. Exploring Big Historical Data: The Historian’s Macroscope. Open Draft Version. Under contract with Imperial College Press. http://themacroscope.org.
Lazer, David, Alex Pentland, Lada Adamic, Sinan Aral, Albert-László Barabási, Devon Brewer, Nicholas Christakis, et al. 2009. “Computational Social Science.” Science 323 (5915): 721–23. https://doi.org/10.1126/science.1167742.
Radford, Alec, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, and Ilya Sutskever. 2022. “Robust Speech Recognition via Large-Scale Weak Supervision.” arXiv. https://doi.org/10.48550/arXiv.2212.04356.
Wang, Chengyi, Sanyuan Chen, Yu Wu, Ziqiang Zhang, Long Zhou, Shujie Liu, Zhuo Chen, et al. 2023. “Neural Codec Language Models Are Zero-Shot Text to Speech Synthesizers.” arXiv. https://doi.org/10.48550/arXiv.2301.02111.
Wu, Yuxin, Alexander Kirillov, Francisco Massa, Wan-Yen Lo, and Ross Girshick. 2019. Detectron2. Meta Research. https://github.com/facebookresearch/detectron2.